Physics of the Earth and Planetary Interiors 261 (2016) 98–106

Contents lists available at ScienceDirect

Physics of the Earth and Planetary Interiors

journal homepage: www.elsevier .com/locate /pepi

Influence of the impoundment of the Three Gorges Reservoiron the micro-seismicity and the 2013 M5.1 Badong earthquake(Yangtze, China)

http://dx.doi.org/10.1016/j.pepi.2016.06.0070031-9201/� 2016 Elsevier B.V. All rights reserved.

⇑ Corresponding authors.E-mail address: hzhang@ucas.ac.cn (H. Zhang).

Huai Zhang a,⇑, Huihong Cheng a,⇑, Yajin Pang a, Yaolin Shi a, David A. Yuen b

aKey Laboratory of Computational Geodynamics, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Beijing 100049, ChinabDepartment of Earth Sciences, College of Science and Engineering, University of Minnesota, 310 Pillsbury Drive SE, Minneapolis, MN 55455-0231, USA

a r t i c l e i n f o

Article history:Received 25 February 2016Received in revised form 17 June 2016Accepted 19 June 2016Available online 22 June 2016

Keywords:Micro-seismicityThree Gorges Reservoir (TGR)Reservoir-triggered-earthquake (RTE)Coulomb failure stress (CFS)Badong earthquake

a b s t r a c t

On December 16, 2013, right after the Three Gorges Reservoir (TGR) reached its highest annual waterlevel, a powerful M5.1 earthquake occurred in Badong County, China’s Hubei Province. The epicenter is5.5 km away from the upstream boundary and 100 km from the dam. Was this earthquake triggeredby the impoundment of the TGR, and what are its subsequences? To answer these questions, we con-structed a coupled three-dimensional poroelastic finite element model to examine the ground surfacedeformation, the Coulomb failure stress change (DCFS) due to the variation of elastic stress and porepressure, and the elastic strain energy potential accumulation in the TGR region upon the occurrenceof this event. Our calculated maximum surface deformation values beneath the TGR compare well withGPS observations, which validates our numerical model. At the hypocenter of the earthquake, DCFS isaround 8.0 � 11.0 kPa, revealing that it may be eventually triggered by the impoundment. We also dis-covered that the total elastic strain energy potential accumulation due to the impounded water load isaround 1.7 � 1012 J, merely equivalent to 0.01% of the total energy released by this event, indicating thatthis earthquake is predominately controlled by the typical regional tectonic settings as well as the weakfault zones, and the reservoir impoundment might only facilitate its procedure or occurrence.Furthermore, the stress level in this region remains high after this earthquake and the subsequentreservoir-triggered micro-seismicity or even bigger event are highly possible.

� 2016 Elsevier B.V. All rights reserved.

1. Introduction with a total water surface area of 1045 square kilometers from

Active debates were raised in recent years whether the ThreeGorges Reservoir (TGR) could trigger big earthquakes. The occur-rence of the 2013 M5.1 Badong earthquake that struck the TGRarea on December 16, 2013, seems to have confirmed the hypoth-esis. Scientists around the world are now wondering the possibilityof subsequent micro-seismicity or even bigger events triggered bythis largest manmade structure, and how the hazard mitigationcan be alleviated if they come to truth in near future. In this studywe carry out a detailed numerical analysis using a coupled three-dimensional poroelastic finite element method to quantitativelyevaluate this open problem.

TGR, the world’s largest power station in terms of the installedpower capacity (22,500 MW), was completed on Yangtze River inJuly 2012. The reservoir is 660 km in length, 1.12 km in width and185 m in depth. Its water storage capacity reaches 39.3 � 109 m3

the western part of Hubei province to the eastern part of Chongqingmunicipality (Mao et al., 2008) (See the upper left panel in Fig. 1).Because of its capacity and typical geological settings, manyresearchers have considered its seismicity risks related to the regio-nal stress state, rock stratum, faults and fracture systems (Chen,2002). Broadband seismic stations were densely deployed to mon-itor the seismicity immediately upon its completion (Che et al.,2010; Ma et al., 2010). However, because of lacking of adequatedata, the results of earlier research were controversial and incon-clusive. For examples, some believed that the micro-seismicactivities might gradually weaken or even disappear over time,while others argued that regional earthquakes would not stop,and the magnitude of earthquakes happened in the reservoir-head region (Dam-Miaohe section) and in the head-mid-reservoirregion (Jiuwanxi-Fairy mountain section and Zigui-Badong section)would be Ms 6 5.0 andMs 6 5.5, respectively (Liao et al., 2009; Sunet al., 1996).

After thefirst impoundment of the TGRon June7, 2003, thewaterhead raised sharply. There were over 7000 micro-earthquakes

Fig. 1. The geological structure and the distribution of the main faults around the TGR. I–IV represent Shenongjia block with Pre-Sinianage, Zigui block with Jurassic age,Huangling block with Pre-Sinianage, and other area with Triassic-sinianage, respectively.①–⑨ denoteWudouhe Fault, Xinhua Fault, Xingshan-Maliang Fault, Fairy mountainFault, Tianyangping Fault, Gaoqiao Fault, Niukou Fault, Shuitianba Fault and Jiuwanxi Fault, respectively. The upper left panel shows the Yangtze River and the location of theThree Gorges Reservoir. The shadow indicates the TGR region.

H. Zhang et al. / Physics of the Earth and Planetary Interiors 261 (2016) 98–106 99

occurred in the upstream area (Chen et al., 2007b). The magnitudesof these events were mostly less than ML2.0, with their distributionpattern closely correlated with the water coverage and water levelfluctuations, indicating that they were essentially reservoir trig-gered events. Moreover, two big earthquakes with ML3.7 andMs4.1 occurred in the reservoir-head region on September 27 andNovember 22, 2008 during its third impoundment (Che et al.,2009, 2010), and theywere followedby a powerfulM5.1 earthquakein Badong County on December 16, 2013, right after the water levelof the TGR reached its annual highest record. The epicenter of thisevent was 5.5 km away from the upper stream boundary and100 km from the dam. The focal depth of this earthquake was5 km, and the focal mechanism revealed that it was a right-lateralnormal dip-slip fault. Later on, two major events with magnitudesof M4.3 and M4.7 also occurred in Zigui County on March 30,2014. These two events are also closely associated to the impound-ment of the TGR.

In this paper, we adopt a fully coupled poroelastic theory (Riceand Cleary, 1976) to evaluate both the instantaneous change ofelastic stress caused by reservoir water load itself, and the subse-quent pore pressure change and the corresponding stress changein the crust beneath the TGR and its vicinity. By taking full account

of the observational data, including the high resolution topogra-phy, water level fluctuations and water depth variation history,flood zone boundaries and fault geometry parameters, we attemptto explain the triggering mechanisms of M5.1 Badong earthquakethrough adopting high resolution finite element modeling andquantitative analysis. In the following sections, we explain thefinite element model of the TRG. This model enables us to flexiblyand efficiently handle different kinds of boundary conditions andcomplex geometries. The model also allow us to quantitativelyestimate the ground deformation for model verification againstavailable GPS observations, the Coulomb failure stress change(DCFS) at the hypocenter for evaluating the earthquake triggeringpotential, and the total accumulated strain energy by water loaditself. And finally, discussions and concluding remarks are givenin the last section.

2. Geological and geophysical settings of the TGR

The occurrence of RTE is closely related to the background tec-tonic stress field, fault-fracture system and rock strength. Thus it isnecessary to consider the tectonic setting of the model (Shi et al.,

100 H. Zhang et al. / Physics of the Earth and Planetary Interiors 261 (2016) 98–106

2013). In the TGR region, fold and fault structures are well devel-oped. There are two large tectonic units in this region. One is theQinling fold belt on the north, the other is the Yangtze platformon the south (Sun et al., 1996). Most of the TGR area is locatedwithin the Yangtze platform, which in turn is subdivided into foursmaller units, known as the Huangling block, the Shennongjiablock, the Zigui block and the faulted block of the Hubei Province,respectively (Li et al., 2009; Sun et al., 1996). Seismic tomographyshows high velocity anomalies down to 20 km beneath the Huan-gling block, and partially molten substances of the upper mantlebeneath the Badong-Zigui block and its adjacent regions (Li et al.,2009; Zhao et al., 2007). Chen et al. (2007a) presented that theelastic wave speed ratio (Vp/Vs) increased before the occurrenceof micro-earthquakes and decreased during the dilatation-diffusion phase.

Three kinds of rocks, granite, limestone and detrital stone, areexposed in the TGR area (Chen, 1999). The first segment of TGR area,from Sandouping toMiaohe, which belongs toHuangling blockwithPre-Sinianage complexes, consists predominantly of strong gran-odiorite with low permeability. The second segment, from Miaoheto Fengjie, which passes through the vicinity of Huangling block,is formed by Pre-Sinian and Upper Cambrian lime stone and detritalrock with higher permeability. Finally, the third segment, upstreamFengjie, is mainly composed of carbonatite, sandstone and shale ofTriassic age (Mao et al., 2008;Wang et al., 2013). The Badong regionis located at the junction of the Huangling block and the Shen-nongjia block. Faults and fractures are well developed understrongly compressive stress state (Han and Rao, 2004).

Three groups of fractures and faults, mostly distributed in theeast of TGR area, can be recognized, as shown in Fig. 1. TheNWW fault-group, including the Fairy Mountain fault, theTianyangping fault and the Wuduhoe fault, extending tens of kilo-meters with a trend of N10� � 40�W. The NNE fault-group, includ-ing the Jiuwanxi fault, Xinghua fault and Niukou fault, ranges fromtens of kilometers to 100 km with a trend of N20� � 30�E. The NEEfault-group, covering the Gaoqiao fault, extending over 40 km witha trend of N450� � 60�E. These three major fault groups dominatethe regional geological settings and crustal deformation, as well asthe nucleation of moderate earthquakes. Ma et al. (2010) proposedthat the micro-earthquakes concentrating around the northern endof Fairy Mountain fault and Jiuwanxi fault, were probably due tothe infiltration of reservoir water into the crust underneath.

3. Methodology

3.1. Coulomb failure stress change (DCFS)

The concept of DCFS is a widely accepted tool to analyze themechanism of earthquake triggering (Stein et al., 1997, 1992). It isalso used in the analysis of tidal triggering (Heaton, 1975;Perfettini and Schmittbuhl, 2001) and reservoir triggering (Chenget al., 2012; Gahalaut and Hassoup, 2012; Ge et al., 2009). As limitedby lacking of reliable measurement techniques of full stress inearth’s crust, the tectonic stress field of the crust beneath the reser-voir cannot be determined in advance. In this study, we introducethe Coulomb failure stress change of Mohr-Coulomb theory andthe Hook-Brown strength criteria to quantitatively investigate theinfluence of the reservoir on the earthquake triggering (Gahalautand Hassoup, 2012; King et al., 1994; Stein et al., 1992). On an exist-ing fault plane, the DCFS is the derivative of the Coulomb stress,which can be expressed as

DCFS ¼ Dsþ lðDrn þ DpÞ þ Dlðrn þ pÞ; ð1Þ

where, Drn is the normal stress change, Ds represents the shearstress change along the slip direction, p and Dp denote the pore

pressure and its change, and l defines the inner friction coefficient.The third term in the right-hand side may be negligible since thereis little experimental data for Dl during the fault-slip process. Eq.(1) is simplifies as

DCFS ¼ Dsþ lðDrn þ DpÞ: ð2ÞAccording to Eq. (2), the fault is prone to slip if DCFS is positive.

On the other hand, if DCFS is negative, the fault becomes morestable. Thus, in principle, we can determine from the calculatedDCFS if an existing fault becomes safer or more dangerous assum-ing that the regional stresses are in a critical state. If the faultgeometry is unknown, we could calculate the optimal DCFS atevery nodal point in the 3-D space.

3.2. Fully coupled poroelastic theory

In this study, we adopt the poroelastic theory of Biot (1941) andRice and Cleary (1976) to obtain the stress change under theimpoundment and drainage of the reservoir. The crust beneaththe TGR is considered as water saturated continuous porous mediaunder quasi-static state. Thus, the governing equation in terms ofthe elastic stress rij and pore pressure p, obeys the equilibriumequation as

rij;j þ f i ¼ 0; ð3Þwhere, rij denotes the stress tensor and f i represents the body force.The constitutive law of continuous poroelastic model is given by

rij ¼ 2Geij þ t1þ t

rkkdij � 3ðtu � tÞBð1þ tÞð1þ tuÞpdij ð4Þ

in which, eij is the strain tensor, p is the pore pressure in solid por-ous rock, and rkk ¼ 1=3rkidik is the effective bulk stress, respec-tively. The relationship between rkk and the bulk strain ekk in asolid structure is rkk ¼ 3Kekk, where dik is the Krönecker delta ten-sor, and G denotes the shear modulus. t and tu represent the Pois-son’s ratios when the material is deformed under drained andundrained conditions, respectively. B is the Skempton’s coefficient(Kuempel et al., 1991; Roeloffs and Rudnicki, 1984; Skempton,1954), which usually is allowed to vary between 0.5 and0.9 � 0.95, depending on the types of the crustal rocks (Roeloffs,1996; Wang et al., 2012). As aforementioned, there are three kindsof rocks, i.e., granite, limestone and detrital stone, in our study area,for simplicity, we choose the value of the B as 0.8.

As to the porous solid material, the fully coupled governingequation satisfies

r2 � rkk þ 3Bp

� �� 1

c@

@trkk þ 3

Bp

� �¼ 0: ð5Þ

Here, c is the diffusion coefficient (m2/s) and can be calculatedfrom

c ¼ kg

2Gð1� tÞ1� 2t

� �B2ð1� 2tÞð1þ tuÞ29ðtu � tÞð1� tuÞ

" #; ð6Þ

where, g is the fluid rheology in a porous solid fault zone.

3.3. Finite element model of the TGR

The occurrence of RTE is closely correlated with the geologicalsettings, regional tectonic stress state, reservoir water level fluctu-ations and physical properties of the rocks and faults (Shi et al.,2013; Simpson et al., 1990). To capture these physical phenomena,we construct a full 3D finite elementmodel of the TGRwith the dataof realistic geometry of the fault system, as shown in Fig. 2. Thedomain of our numerical model is 110.0 � 111.5�E, 30.3 � 31.8�Nin horizontal. The vertical depth is from the surface down to

Fig. 2. The high-resolution finite element model of TGR. I–IV represent the geological blocks, and ①–⑨ show faults in the TGR region the same as in Fig. 1.

H. Zhang et al. / Physics of the Earth and Planetary Interiors 261 (2016) 98–106 101

�20 km. The width of the fault zone is less than 7 km. At the sametime the TGR locates in a canyon with elevation variation of hun-dreds of meters between the water level and the local mountains.It is obvious insufficient if the topographic effect is not considered(Cao and Shi, 2011). Thus, in this study, we consider the topographyin our model. The whole model consists of 17 sub-layers, with968,949 triangular prismatic type of finite elements composed of518,886 nodes. The total simulation period is from June 2003, whenthe first impoundment of the TGR was getting started, until May2014. The time step is one month, and the total time steps are132. From the assumption of poroelastic theory, the initial porepressure is assumed as hydrostatic pressure in water-saturatedstate of equilibrium. The pore pressure on top surface of thereservoir’s bedrock is assigned as of the water head from the waterdepth impounded, and the pore pressure on both the lateral and thebottom surfaces are given by a hydrostatic pore pressure with the

Table 1The values of physical parameters in our finite element model.

Block Depth (km) Model 1 Model 2

E (1010 P

SNJ 0–2 6.582–5 6.945–11 7.6211–20 9.19

ZG 0–2 4.972–5 5.225–11 6.9911–20 7.64

HL 0–2 8.472–5 8.345–11 9.4611–20 10.1

OTHER 0–2 E (1010 Pa) = 6.0m = 0.25C (m2/s) = 0.2

6.582–5 6.555–11 8.1011–20 9.19

FAULT 3.00

depth. The normal displacement and the tangential shear stresson the lateral boundaries of the whole model are fixed. While, thenormal stress and pore pressure of the reservoir bottom areassigned as hydrostatic pressure varying with the water levelfluctuations during the impoundment and drainage processes.

Hydraulic parameters, especially the diffusion coefficient, areclosely related to the pore pressure and varying with respect totime in poroelastic problems. Unfortunately, there are few hydrau-lic parameters data of TGR area available. Here, according to themethod of Talwani et al. (2007), we calculated the seismicdiffusion coefficients of different subdomains based on the seismicactivities and seismic stagnant time windows after the impound-ment of TGR (Xu, 2010). The average diffusion coefficient is0.2 m2/s and its maximum value is 2.0 m2/s. For comparison, weassign Model 1 as homogeneous medium with diffusion coefficientas 0.2 m2/s. Because of the high diffusion of fault and Zigui block

a) m Vp Vp/Vs C (m2/s)

0.23 5.50 0.35 0.20.27 5.76 0.320.25 5.87 0.330.26 6.32 0.320.23 4.97 0.35 2.00.27 5.20 0.320.25 5.69 0.330.26 5.92 0.320.23 6.02 0.35 0.20.27 6.15 0.320.25 6.34 0.330.26 6.53 0.320.23 5.50 0.35 0.20.27 5.64 0.320.25 6.00 0.330.26 6.32 0.320.22 3.00 – 2.0

102 H. Zhang et al. / Physics of the Earth and Planetary Interiors 261 (2016) 98–106

with distributed carbonate rock, Model 2 is assigned as inhomoge-neous medium, the fault systems and Zigui block designated as2.0 m2/s. At the same time, the selection of elastic parameters arebased on the in-situ geological survey of Vp and Vp/Vs ratio of dif-ferent depths (Li et al., 2009, 2010; Zhao et al., 2007), the detailedelastic parameters are listed in Table 1 (Simpson et al., 1990).

4. Simulation results

4.1. Deformation induced by the TGR

Since 1998, advanced dense GPS network has been deployed tomonitor the horizontal and vertical displacement of the TGR area.Later on, through numerical and analytical models of homoge-neous materials, researchers examined the elastic displacementof the TGR and its adjacent regions under the gravitational loadof reservoir water (Liang and Hu, 2008; Wu et al., 2009). In orderto verify our 3D numerical model, we first calculate the displace-ment of the TGR area when the water level is 135 m, as shown inFig. 3(a). The maximum subsidence displacement is �35 mm,which is in good agreement with the GPS measurement by Liangand Hu (2008), confirming that our numerical model results arereliable. At the same time, the subsidence is mainly concentratedin the center of the reservoir at the Xiangxi segment and graduallydecreases to zero upon the boundaries of the reservoir. The defor-mation of the Zigui Block, where the TGR passes through, is rela-tively bigger than that of other areas. The surface deformationnear the Badong area is 5 mm. Fig. 3(b) demonstrates the verticaldisplacement when the water level reaches 175 m. We see themaximum surface deformation is up to 70 mm. But in the Badongarea, the maximum surface deformation is �10 mm.

4.2. Temporal variations of stress at the hypocenter

After the impoundment of the TGR, many micro-earthquakesoccurred around the reservoir region and were concentrated inthe Badong area and the Fairy mountain-Jiuwanxi fault zone,exhibiting linearized or mass-like clustering patterns, as shownin Fig. 1. Therefore, herein, we not only focus on the M5.1 earth-quake in the Badong region, but also analyze the Ms4.1 earthquake

Fig. 3. Numerical calculation of the vertical displacement distribution after the impoundmtwo focal mechanism beach ball diagrams represent the epicenters of the M5.1 earthq(right).

on November 22, 2008, which have been seen as the second big-gest RTE by the TGR, as comparison.

Fig. 4(a) shows the pore pressure and DCFS at the hypocenter ofthe M5.1 earthquake. Both of them increase gradually along withthe water level arising, and the increase of pore pressure is muchlarger than the elastic stress after the impoundment of the TGR.The DCFS caused by the TGR water load is close to 1.5 kPa whenthe M5.1 earthquake occurred, indicating that the water load lever-aged the risk of fault activation and facilitated the fault plane torupture. More strikingly, when pore pressure effect is considered,DCFS can reach up to 8.0 � 11.0 kPa when the friction coefficientis 0.6, indicating that the effective pore pressure is relatively signif-icant. Most probably, pore pressure predominated the triggeringmechanism of this seismogenic faults towards destabilization.

By comparing the homogeneous model (Model 1) with theinhomogeneous model (Model 2), we get: (1) After the initialimpoundment of the TGR, the pore pressure change of Model 1 isbigger than that of Model 2, which is calculated with higher diffu-sion coefficient. This phenomenon suggests that the undrainedresponse is instantaneous, and its subsequent diffusion effectshould not to be neglected, especially when its diffusion coefficientis low. For the undrained response, pore pressure change is owingto the elastic bulk deformation of the solid porous skeleton. Mean-while, the additional water load also leads to elastic deformationand stress of the solid porous skeleton. Combining the instanta-neous effects of undrained response, and elastic stress of the solidporous skeleton by water load, we can explain why some micro-earthquakes appeared around the reservoir area immediately afterthe onset of the impoundment. (2) With the time passing, addi-tional water head escalates the pore pressure diffusing throughthe faults and fractures beneath the water reservoir (diffusiveresponse), and increases the pore pressure at the hypocenter grad-ually. Apparently, during this procedure, the crust’s diffusion coef-ficient exhibits a predominant effect on the total pore pressurechanges at the hypocenter. This is the mechanism of which thepore pressure of Model 2 exceeds that of Model 1 eventually.

The Ms4.1 earthquake, occurred on the Fairy mountain thrustfault, was the first relative big event after the water level of theTGR reached 175 m. Its epicenter is nearly 1.0 km from the bankof TGR and less than 30 km from TGR dam. Fig. 4(b) shows the vari-ation of pore pressure and DCFS at its hypocenter. Comparing the

ent of TGR when the water level reached 135 m (a) and 175 m (b), respectively. Theuake in the Badong region (left) and the Ms4.1 earthquake on November 22, 2008

Fig. 4. Variation of pore pressure and Coulomb failure stress change at thehypocenter in terms of the TGR water level fluctuation. (a) DP and DCFS at thehypocenter of the M5.1 Badong earthquake. The parameters of the correspondingseismogenic fault planes are referenced by the previous geological field investiga-tion, for examples, the strike angle is 74�, the dip angle is 56�, and the rake angle is�178�. (b) DP and DCFS of the hypocenter at the Ms4.1 earthquake .The calculationparameters of seismogenic fault are: strike angle = 213�, dip angle = 81� and rakeangle = 110� (Xu, 2010).

H. Zhang et al. / Physics of the Earth and Planetary Interiors 261 (2016) 98–106 103

Ms4.1 earthquake to the M5.1 earthquake, although both of thesetwo earthquakes occurred when water level reached the highestannual water level, they were significantly different. Firstly, theDCFS of the Ms4.1 earthquake’s hypocenter induced by the waterload is negative while positive of M5.1 earthquake, suggesting thatthe impoundment of TGR actually stabilized the faults and pre-vented the hypocenter from rupturing when the occurrence ofMs4.1 earthquake. Given this, we can preliminarily conclude that,if the reservoir is located above the hanging wall of the steeply dip-ping reverse fault or shallowly dipping thrust fault, the impound-ment of the reservoir makes the thrust fault more stable, butmay lead the normal fault to be more dangerous. On the contrary,the drainage of the reservoir lead the normal fault to more stable,but makes the thrust fault may be more dangerous. That is to say,the effect of reservoir impoundment depends on where it is locatedand the nature of the local seismogenic fault system. Secondly, asfor the Ms4.1 earthquake, when the pore pressure effect is consid-ered for the homogenous model, the DCFS can only reach up to�1.0 kPa in model 1, revealing that the impoundment of the TGRmay not trigger the occurrence of earthquake when adopting thehomogenous model. However, when adopting the heterogeneousmodel, the maximum DCFS at the hypocenter is around 5.0 kPa.These results confirm that the computational values are highlysensitive to RTE model parameters, especially, the values of the dif-fusion coefficient. When the geological background is considered,the calculated results could be consistent with the actual situa-tions. This means, it is very necessary to adopt high-resolutionheterogeneous geological model to analyze the realistic effects ofreservoir loading in future studies. Moreover, DCFS value at thetime of the impoundment may not be a good indicator of RTE

and the evaluation of earthquake risk must consider the long-term effect due to pore pressure diffusion.

Furthermore, by comparing the difference between these twoearthquakes, their patterns of the pore pressure increasing areslightly different, as shown in Fig. 4. The Ms4.1 earthquake doesnot locate in the Zigui block with high diffusion coefficient, thepore pressure of model 2 is always larger than model 1. It confirmsthat their triggering mechanisms are slightly different too.

More generally, at the beginning of the impoundment,undrained and elastic effects are significant in response to the fastwater load increasing. The diffusive response gradually becomepredominant over time and may overtake the undrained and elas-tic effects eventually, mainly depending on the permeability of thecrust as well as the nature (normal/thrust/reverse) and geometryof the seismogenic fault systems beneath the reservoir. Duringthe impoundment or drainage of the reservoir, the tradeoffbetween undrained and diffusive effects constitutes a process ofdynamic responses. More specifically, when the diffusion coeffi-cient is small, undrained response is more obvious, and when thediffusion coefficient increases, diffusive response graduallyincreases with respect to time. Essentially, at the hypocenter ofthe seismogenic fault, whether the undrained response or diffusiveresponse is predominant depends on the fault geometry, the focaldepth and permeability of the crust beneath the reservoir.

Accordingly, we come to the preliminary conclusions that: (1)During initial impoundment or drainage process, while the changein water level is fast, the micro-earthquakes are more elastic andundrained response related. The magnitudes of these earthquakesare small and their occurrences are relatively frequent and exhibita spatially intensive pattern. (2) After the curst undergoes with thediffusive effect over a long time, the triggered seismicity is finallyattributed to the delayed dynamic process of pore pressure redis-tribution under the water head. And typically, this kind of triggeredseismicity is spatially and temporally sparse, and its magnitude isrelatively larger.

4.3. Stress changes at different depths with water level fluctuation

The water level of the TGR underwent from 135 m in 2003 to156 m in September 2006, and reached up to its highest value of175 m in November 2010, and then fluctuated with the seasons.For focal depths of most micro-RTES were around 5 km (Ma et al.,2010), Fig. 5(a–c) demonstrate the pore pressure change (Dp) atthe depth of 5 km of these three time windows mentioned above.As the diffusion of the pore pressure Dp obviously increases withtime, the pore pressure changes at the reservoir center area andgradually spread around. At the beginning of the impoundment,Dp between Miaohe and Xiangxi evidently and rapidly increases,and then Dp from Xiangxi to Badong increases at a larger rate afterNovember 2010. But due to the different geometries of the Gaoqiaofault and Fairy mountain fault, the seismicity around Badong areaincreases more faster than that of the Zigui area at the beginningof impoundment. These results are closely correlated with the seis-mic activities of the Zigui area, whose reservoir seismicity, owing toits high permeability, become more intensive with respect to time.

Fig. 5(d–f) clearly reveal that DCFS varied with different depthsin December 2013. In most river channels, DCFS reaches up to65.0 kPa at the depth of 3 km, while only exceeding 20.0 kPa atBadong, the Zigui Block and the segment of Xiangxi to Miaohe atthe depth of 5 km, respectively. The DCFS attenuates rapidly withthe depth, only slightly larger than 20.0 kPa at the depth of 11 kmin the Xiangxi to Zigui segment. Comparing the distribution pat-tern of 119 micro-earthquakes (Xu, 2010), that occurred fromMarch 2009 to December 2009, we can infer that most of theseearthquakes are well located in the regions with positive DCFS,suggesting that the micro-earthquakes occurred around the TGR

Fig. 5. The distribution of the pore pressure changes at the depth of 5 km with water level, and Coulomb failure stress changes varies with different depths during theoccurrence of Ms5.1 earthquake in model 2.The inner frictional efficient is 0.6, the strike angle is 74�, the dip angle is 56�, and the rake angle is �178�, respectively. The focalmechanisms of the micro-earthquakes are adopted from (Xu, 2010). The magnitude of these 119 micro-earthquakes are larger than 1.0 from March 2009 to December 2009.Among them, the focal depths of 59 micro-earthquakes are less than 4.0 km in Fig. 5(d), the focal depths of 43 micro-earthquakes are less than 7.0 km and larger than 4.0 kmin Fig. 5(e), and the focal depths of 17 micro-earthquakes are larger than 7.0 km and less than 11.0 km in Fig. 5(f).

104 H. Zhang et al. / Physics of the Earth and Planetary Interiors 261 (2016) 98–106

have direct correlation with the impoundment of the reservoir. Atthe same time, most of micro-earthquakes had shallow depth (lessthan 6 km) and were located in the Badong area.

4.4. Strain energy potential accumulation induced by the TGR

After the impoundment of the TGR, strain energy potential wasaccumulated in the crust beneath the reservoir and its adjacentregions. For the sake of analyzing the effect of the reservoirimpoundment rate on the earthquake occurrence, the elastic strainenergy potential accumulation induced by the reservoir is calcu-lated. Basically, the elastic strain energy potential is the integrationof elastic strain energy density function over spatial and temporaldomains (Li et al., 2010) as

U ¼Zvwðx; y; z; tÞdvdt ¼

Zv

12rijeijdvdt; ð8Þ

where, wðx; y; z; tÞ is the strain energy density rate. Based on theestimation of our numerical simulation results for the M5.1 earth-quake, the mechanical strain energy accumulation of the entireTGR and its adjacent regions (about 5.5 � 104 km3) is around1.7 � 1012 J. However, according to the estimation criterion of theseismic wave energy released by the seismic waves (Bath andDuda, 1964), as shown by

log E ¼ ð12:24� 1:35Þ þ ð1:44� 0:20ÞM; ð9Þthe seismic wave energy released by the M5.1 earthquake is about1019 J. Thus, the contribution of the mechanical strain energy

induced by the TGR is merely less than 0.01% of the total energyreleased by the M5.1 earthquake, signifying that the impoundmentof the reservoir only triggered, rather than providing the strainenergy for this RES.

5. Discussions

Reservoir-triggered-earthquake (RTEs) is one of the significantissues in the earthscience community (Kerr and Stone, 2009; Geet al., 2009). Worldwide, there have been more than 140 RTEsreported by over 70 countries (Mao et al., 2008; Talwani et al.,2007). Among them, 4 reservoir regions had been stricken by RTEswith magnitudes MP 6.0. Most of the existing investigations con-firm that there is a close correlation between the reservoir waterlevel fluctuation and seismicity (Chen and Talwani, 1998; Gupta,1985; Wu et al., 2009). The current understanding of the mechani-cal processes following reservoir impoundment suggests twomajorphases. First, a rapid change of elastic stress coupled withundrained pore pressure dissipation, owing to the bulk deformationof the solid skeleton of the crust under the load of the water body.This leads to quick response of weak faults with planes oriented at acritical angle (Simpson et al., 1988; Talwani, 1997). An example ofthis type of response is the many shallow micro-earthquakesoccurred immediately after the impoundment of the Kariba Reser-voir since its impoundment in December 1958 (Gough and Gough,1970a,b). Second, the pore pressure state in the crust beneath thereservoir and its adjacent regions changes due to its diffusive effect.Consequently, not only the undrained pore pressure in the fracture-

H. Zhang et al. / Physics of the Earth and Planetary Interiors 261 (2016) 98–106 105

fault system gradually dissipating into its surroundings, but alsothe pore pressure at the focal depth gradually increases due to dif-fusion from the reservoir bed by the water head. This could lead tolarge earthquakes after a certain period of time following theimpoundment. For example, theML5.7 Aswan Reservoir earthquakeoccurred nearly 17 years after its first impoundment (Gahalaut andHassoup, 2012). These two physical processes indicate that thestress change at the hypocenter can rise up rapidly from a coupledelastic response due to compaction of the porous space (elasticdeformation and undrained response), or more slowly with the dif-fusive effect of water head from the reservoir’s bed (diffusiveresponse). Overall, the stress change at the hypocenter or on a faultplane is closely associated with the coupled effects of all the differ-ent mechanisms mentioned above. According to the theory of rockmechanics, changes in either the elastic stress (decreased normalstress or increased shear stress) or decreased effective normalstress due to increased pore pressure may facilitate the failure offaults with a critical orientation.

Heated arguments occurred after the 2008 Ms8.0 Wenchuanearthquake on whether the impoundment of the Zipingpu reser-voir, which is located close to the LongmenShan fault zone, trig-gered the occurrence of this huge devastating event (Ge et al.,2009; Kerr and Stone, 2009, 2010). Many scholars have continuedto study the tectonic activities and seismicity beneath this reser-voir in relation to the earthquake. Reviews of the more recent stud-ies were proposed by Ma and Zhou et al. (Ma et al., 2012; Zhouet al., 2012).

Upon the occurrence of the M5.1Badong earthquake, the micro-earthquakes with magnitudes less than M2.0 became more active.Moreover, comparing to the Ms6.1 earthquake happened inXinfengjiang Reservoir in 1962, which is well acknowledged asthe largest RTE in China so far, the DCFS we obtained at itshypocenter is about 3.0 kPa (Cheng et al., 2012). The primaryresults of our simulations suggest that the DCFS of the hypocenterof theM5.1 Badong earthquake varies from 8.0 to 11.0 kPa, suggest-ing that the impoundment of the TGR mostly possibly triggered theoccurrence of this event. On the other hand, because the regionaltectonic stress background upon this event is still not clear andthe DCFS we obtained here is still too small. We cannot rule outthe possibility of model uncertainty, for examples, fault geometricparameters, boundary conditions and initial conditions of porepressure. So it cannot be directly concluded that there must be adirect connection between the TGR impoundment and the occur-rence of this earthquake. However, because the GPS observationof crustal motion of the Xinfengjiang reservoir area is larger thanthat of the TGR (Gan et al., 2007), tectonically, the TGR region isnot as active as that of the Xinfengjiang reservoir region. So, maybethe magnitudes of future earthquakes triggered by the impound-ment of TGR are less than that of the 1962 Xinfengjiang earthquake.

In this study, at the hypocenter, the mechanism of elastic stresschange induced by the water load is different in accordance withthe different typical characteristics of seismogenic faults. In thecase of a thrust fault, the elastic stress change from the impound-ment of reservoir usually makes it more stable, whereas the nor-mal fault may become more dangerous. The seismogenic faults ofthe November 2008 M4.1 and the Wenchuan earthquakes are boththrust types of faults, and their DCFSs induced by the elastic stresschange are all negative. However, DCFSs induced by the elasticstress change of the December 2013 M5.1 and XinfengjiangMs6.1 events are all positive.

6. Conclusions

The DCFS at the hypocenter of the M5.1 Badong earthquake ispositive at the impoundment of the reservoir. Both pore pressure

and the DCFS at the hypocenter gradually increased with time.At the occurrence of the earthquake, pore pressure at the hypocen-ter reaches 13.0 � 17.0 kPa and DCFS is around 8.0 � 11.0 kPa,which is above the threshold value of stress changes for the occur-rence of reservoir triggered earthquake, as proposed by King et al.(1994). Thus the occurrence of the earthquake could be due to theimpoundment of the TGR. On the other hand, the DCFS at thehypocenter of the M4.1 earthquake in November 2008 in the ZiguiCountry (Mei et al., 2013) is negative during the impoundment.Diffusion of pore pressure cause the DCFS to gradually increase,reaching up to �5.0 kPa upon the occurrence of the M4.1 earth-quake. These examples show that the DCFS at the time of theimpoundment may not be a good indicator of RTE, and evaluationof earthquake risk must consider the long-term effects due to porepressure diffusion.

Scientists are now wondering if other large earthquakes may betriggered by this large manmade structure and how the relatedhazards may be alleviated. Our simulation shows that the elasticstrain energy released by this earthquake is very small. On theother hand, the stresses continued to change due to pore pressurediffusion, which accounts for the predominant increase of theDCFS. Thus we cannot exclude the possibility of the occurrence ofsubsequent big reservoir-triggered events in the TGR area in thefuture. Long-term simulation of the DCFS on the major faultsbeneath the TGR may be needed to evaluate future seismic risksin this area.

7. Data and resources

The topographical data used in this paper was obtained fromthe SRTM http://dds.cr.usgs.gov/srtm/ (last accessed March2014). The coordinates of hypocenter and focal mechanism solu-tions of the M5.1 Badong earthquake were downloaded from ChinaEarthquake Administration (www.ceic.ac.cn) (last accessedDecember 2013). The details of the Ms4.1 earthquake was fromthe Seismological Bureau of Hubei Province, China. http://www.eqhb.gov.cn/ (last accessed May 2014).

Acknowledgments

This research is partial supported by National Basic ResearchProgram of China under grant number 2014CB845906, andNational Science Foundation of China (41590864, 41274103 and41404078). It is also partially supported by the CAS/CAFEA Interna-tional Partnership Program for creative research teams (No.KZZD-EW-TZ-19 and KZZD-EW-TZ-15).

References

Bath, M., Duda, S.J., 1964. Earthquake volume, fault plane area, seismic energy,strain, deformation and related quantities. Ann. Geofis. Rome 17, 353–368.

Biot, M.A., 1941. General theory of three-dimensional consolidation. J. Appl. Phys.12, 155–164.

Cao, J., Shi, Y., 2011. Numerical study on channel reservoir induced earthquake. J.Grad. School Chin. Acad. Sci. 28, 19–26.

Che, Y., Chen, J., Zhang, L., Yu, J., Liu, C., Zhang, W., 2009. Study of the reservoir-induced Hujiaping Ms4.1 earthquake in the Three Gorges Dam area. Earthquake29, 1–13.

Che, Y., Chen, J., Zhang, L., Yu, J., Liu, C., Zhang, W., 2010. The Ms4.1 reservoir-induced Hujiapping earthquake in the Yangtze Three Gorges Reservoir region.Earthquake Res. China 24, 304–319.

Chen, D., 1999. Engineering geological problems in the Three Gorges Project on theYangtze, China. Eng. Geol. 51, 183–193.

Chen, D., 2002. Summary of researches on reservoir induced seismicity at ThreeGorges. China Three Gorges Constr. 5, 4–11.

Chen, L., Talwani, P., 1998. Reservoir-induced seismicity in China. Pure Appl.Geophys. 153, 133–149.

Chen, J., Gan, J., Li, P., Xue, J., Li, F., Li, Y., 2007a. On wave velocity ratio (Vp/Vs) inBadong area after impoundment of Three Gorges Reservoir. J. Geod. Geodyn. 27(12–14), 42.

106 H. Zhang et al. / Physics of the Earth and Planetary Interiors 261 (2016) 98–106

Chen, J., Gan, J., Li, S., Di, L., Xue, J., 2007b. Seismic activity characteristics inDongrangkou area, Badong country of Hubei Province. J. Geod. Geodyn. 27, 97–105.

Cheng, H., Zhang, H., Zhu, B., Sun, Y., Zheng, L., Yang, S., Shi, Y., 2012. Finite elementinvestigation of the poroelastic effect on the Xinfengjiang reservoir-triggeredearthquake. Sci. China Earth Sci. 52, 1942–1952.

Gahalaut, K., Hassoup, A., 2012. Role of fluids in the earthquake occurrence aroundAswan reservoir. Egypt. J. Geophys. Res. 117, 1–13.

Gan, W., Zhang, P., Shen, Z., Niu, Z., Wang, M., Wan, Y., Zhou, D., Cheng, J., 2007.Present-day crustal motion within the Tibetan Plateau inferred from GPSmeasurements. J. Geophys. Res. 112, 1–14.

Ge, S.M., Liu, M., Lu, N., Godt, J.W., Luo, G., 2009. Did the Zipingpu reservoir triggerthe 2008 Wenchuan earthquake? Geophys. Res. Lett. 36, 1–5.

Gough, D.I., Gough, W.L., 1970a. Stress and deflection in the lithosphere near LakeKariba. Geophys. J. Roy. Astron. Soc. 21, 70–101.

Gough, D.J., Gough, W.L., 1970b. Load-induced earthquakes at Lake Kariba. Geophys.J. Roy. Astron. Soc. 21, 79–101.

Gupta, H.K., 1985. The present status of reservoir induced seismicity investigationswith special emphasis on Koyna earthquakes. Tectonophysics 118, 257–279.

Han, X., Rao, Y., 2004. Analysis of genesis of reservoir-induced microquakes inBadong reach of Three Gorges. J. Geod. Geodyn. 24, 74–77.

Heaton, T.H., 1975. Tidal triggering of earthquakes. Geophys J. R. Astron. Soc. 43,307–323.

Kerr, R.A., Stone, R., 2009. A human trigger for the great quake of Sichuan? Science323, 322.

Kerr, R.A., Stone, R., 2010. Two years later, new rumblings over origins of Sichuanquake. Science 327, 1184.

King, G.C.P., Stein, R.S., Lin, J.P., 1994. Static stress changes and the triggering ofearthquake. Bull. Seismol. Soc. Am. 84, 935–953.

Kuempel, P.L., Henson, J.M., Dircks, L., Tecklenburg, M., Lim, D.F., 1991. Dif, a recA-independent recombination site in the terminus region of the chromosome ofEscherichia coli. New biolog. 3, 799–811.

Li, Q., Zhao, X., Cai, J., Liu, R., Long, G., An, Y., 2009. P wave velocity structure of upperand middle crust beneath the Three Gorges Reservoir dam and adjacent regions.Sci. China, Ser. D Earth Sci. 52, 567–578.

Li, S., He, T., Yin, X., 2010. Introduction to Fracture Mechanics of Rock (in Chinese).China Science and Technology University Press, Hefei, pp. 2–20.

Liang, X., Hu, R., 2008. Numerical simulation of vertical crustal deformation in ThreeGorges area based on SRTM. J. Geod. Geodyn. 28, 38–43.

Liao, W., Zhang, L., Yao, Y., 2009. Characteristics of seismicity in the Three GorgesReservoir area. Seismol. Geol. 31, 707–714.

Ma, W., Ma, C., Li, H., Yuan, J., Xu, X., Zhang, X., Zhang, L., 2010. Intensive observationof reservoir-induced earthquake and preliminary analysis on the causes ofearthquakes in Three Gorges Reservoir. Seismol. Geol. 32, 552–563.

Ma, W., Xu, X., Xu, G., Yuan, J., Xu, C., Lin, Y., 2012. Assess the reservoir-inducedseismic hazard in the Hubei section of the Three Gorges Reservoir using grayclustering method. Seismol. Geol. 34, 726–738.

Mao, Y., Ai, Y., Li, Z., Chen, R., Hu, X., Xie, L., Fu, H., Wu, M., Li, M., 2008. Study ofReservoir-Induced Seismicity. Seismological Press, Beijing, p. 262.

Perfettini, H., Schmittbuhl, J., 2001. Periodic loading on a creeping fault:implications for tides. Geophys. Res. Lett. 28, 435–438.

Rice, J.R., Cleary, M.P., 1976. Some basic stress diffusion solutions for fluid-saturatedelastic porous media with compressible constituents. Rev. Geophys. Space Phys.14, 227–241.

Roeloffs, E., 1996. Poroelastic techniques in the study of earthquake-relatedhydrologic phenomena. Adv. Geophys. 37, 135–195.

Roeloffs, E., Rudnicki, J.W., 1984. Coupled deformation-diffusion effects on water-level changes due to propagating creep events. Pure Appl. Geophys. 122, 560–582.

Shi, Y., Cheng, H., Zhang, B., Zheng, L., Zhang, H., 2013. Are Reservoir EarthquakesMan-Made? WIT Press Water and Society, Boston, pp. 273–284.

Simpson, D.W., Leith, W.S., Scholz, C.H., 1988. Two types of reservoir-inducedseismicity. Bull. Seism. Soc. Am. 78, 2025–2040.

Simpson, D.W., Gharib, A.A., Kebeasy, R.M., 1990. Induced seismicity and changes inwater level at Aswan reservoir. Gerlands Beitr. Geophys. 99, 191–201.

Skempton, A.W., 1954. The pore-pressure coefficients A and B. Geotechnique 4,143–147.

Stein, R.S., King, G.C.P., Lin, J., 1992. Change in failure stress on the southern SanAndreas fault system caused by the 1992 magnitude = 7.4 Landers earthquake.Science 258, 1328–1332.

Stein, R.S., Barka, A.A., Dieterich, J.H., 1997. Progressive failure on the NorthAnatolian fault sinee 1939 by earthquake stress triggering. Geophys. J. Int. 128,594–604.

Sun, Y., Tan, C., Wang, R., Hu, D., 1996. Assessment and zonation of regional crustalstability in and around the dam region of the three Gorges project on theYangtze River. Acta Geosci. Sinica 17, 258–268.

Talwani, P., 1997. On the nature of reservoir-induced seismicity. Pure Appl.Geophys. 150, 473–492.

Talwani, P., Chen, L., Gahalaut, K., 2007. Seismogenic permeability, ks. J. Geophys.Res. 112, 1–18.

Wang, X.Q., Schubnel, A., David, E.C., Guéguen, Y., Ge, H.K., 2012. High Vp/Vs ratio:saturated cracks or anisotropy effects? Geophys. Res. Lett. 39, 6099–6112.

Wang, Q., Zhang, L., Liao, W., Li, J., Shen, X., 2013. Fault tectonics and seismic activitycharacteristics of Three Goreges Reservoir. J. Geod. Geodyn. 33, 29–33.

Wu, J., Chen, S., Chen, J., 2009. Numerical simulation of toral displacement fieldchange after impoundment of the three Gorges reservoir. J. Geod. Geodyn. 29,52–55.

Xu, C., 2010. A Study on Reservoir-Induced Earthquake Location and FocalMechanism Character of Yangtze River Three Gorges Reservoir Region.Institute of Geology, China Earthquake Administration. Graduate School of theChinese Academy of Sciences, Beijing, pp. 1–78.

Zhao, X., Li, Q., Cai, J., 2007. On minimum 1D velocity model applied in three Gorgesreservoir area. J. Geod. Geodyn. 27, 1–7.

Zhou, L., Zhao, C., Chen, Z., Zheng, S., 2012. Three-dimensional VP and VP/VSstructure in the longtan reservoir area by local earthquake tomography. PureAppl. Geophys. 169, 123–139.